Summary: The essential circuit for both item and associative stimulus recognition in any given sensory modality (or across modalities) consists of the relevant cortical sensory processing stream(s), the medial temporal periallocortex (i.e. parahippocampal, perirhinal, and entorhinal cortices), the ventromedial prefrontal cortex, and the magnocellular division of the medial dorsal nucleus of the thalamus. Associative recall, on the other hand, appears now to be organized hierarchically; thus, whereas context-free recall, or fact memory, also seems to depend primarily on the above basic memory circuit, context-rich recall, or event memory, seems to depend in addition on a higher-order circuit superimposed on the basic one and consisting of the hippocampus, mamillary body, anterior thalamic nuclei, and, possibly, cingulate cortex. That item recognition at least does not depend on the higher-order memory circuit is supported by new evidence obtained in Jon (one of the patients with 'developmental amnesia' described initially in Science 277: 376-380, 1997), who has seemingly selective bilateral hippocampal pathology induced by a neonatal hypoxic/ischaemic insult. On standardized memory tests allowing quantitative assessment, Jon's recognition scores were at or above the 50th percentile, whereas his recall scores fell below the 1st percentile. That associative recognition in monkeys also does not depend on the higher-order memory circuit (but does require the basic circuit) is supported by new evidence obtained in monkeys. These results indicate that the ability to form object-place associations is unaffected by selective, excitotoxic damage to the hippocampus, and yet is severely impaired by ablation of the underlying parahippocampal tissue. In the patients with developmental amnesia, however, relatively selective hippocampal damage did seem sufficient to impair memory for object-place associations. One potentially important difference between the two studies is that the patients were required to remember 20 object-place associations, presented as a single list for several successive trials, whereas the monkeys were required to remember only two such associations at a time, with each pair presented just once (i.e. one-trial memory for object-place associations). To approximate more closely the task given the patients, we used a 10-well board to train monkeys in successive steps to learn a set of 10 object-place associations. The final test consisted of randomly pairing each object in its correct (baited) place with a distracter (one of the other nine objects in an incorrect, unbaited, place). After acquisition, half of the animals received excitotoxic hippocampal lesions (Group HC), and half were kept as unoperated controls (Group N). All animals were then tested for retention of the preoperatively learned set and then trained on two additional sets. There was no impairment in Group HC at any stage of the experiment, suggesting that differences other than set size underlie the different outcomes in the human and animal studies. These other differences remain to be explored. We previously reported that hypoxic events sustained perinatally and those sustained in mid childhood can result in the same neuropsychological profile of developmental amnesia, i.e. markedly impaired episodic memory and relatively preserved semantic memory, in association with equivalent extents of hippocampal atrophy. Subsequent morphometric analyses of the whole brain revealed that, compared to age-matched controls, children with this syndrome whose pathology was incurred perinatally had bilateral reduction of grey matter in the putamen and ventral thalamus as well as the hippocampus. We have now compared the results of such analyses in two groups, the one with pathology sustained perinatally, and the other, between the ages of 6 and 13 years. Both showed bilateral abnormality in the hippocampus, putamen, and thalamus, as evidenced by conjunction analysis, which looks for common abnormalities across groups. The only other significant abnormality that was found in both groups by conjunction analysis was an area in the right posterior cingulate gyrus. The similarity of the amnesic and neuropathological patterns in the two groups suggests that, if the form of this memory disorder is a special syndrome related to the earliness of the hypoxia-induced damage, then the effective age at injury must extend from birth to puberty. Compared to age-matched controls, adolescents who were born preterm and required at least one week of ventilation showed a mean reduction in hippocampal volume on each side of approximately 10 percent (range, 0 to 25 percent), whereas adolescents who had suffered early hypoxic-ischaemic episodes showed a mean reduction on each side of approximately 40 percent (range, 30 to 55 percent). Both IQ scores and scores on immediate memory tasks were equivalent in the two patient groups. A wide variety of verbal and visual delay tasks, however, revealed marked group differences. The preterm group was significantly impaired relative to the controls on a few measures only: route following and prospective memory, including remembering an appointment and a belonging. The hypoxic-ischaemic group, by contrast, were impaired relative to both of the other groups on all delay tasks, a finding consonant with the diagnosis of developmental amnesia. Discriminant analysis indicated that the Rivermead Behavioural Memory Test classified the children into groups as accurately as the hippocampal volumes. Interestingly, on one non-delay measure, the numerical operations subtest of the Wechsler Objective Numerical Dimensions, the group with developmental amnesia was significantly superior to the preterm group. The results suggest that early hippocampal pathology leads to developmental amnesia only when the bilateral volume of this structure is reduced approximately 25-30 percent below normal on each side. To determine the extent to which the visual recognition impairment produced by perirhinal ablations might be due to subcortical fiber damage, we injected the neurotoxin, ibotenate, bilaterally into the perirhinal cortex of monkeys that had been trained in the rule for delayed nonmatching-to-sample (DNMS) with trial-unique objects. The postoperative performance of these animals (Group IBO) was compared with that of two previously studied groups, one with perirhinal aspiration lesions (Group ASP) and the other an intact control (Group NC), with an N of 4 in each group. The average relearning scores of Group IBO did not differ reliably from those of Group ASP, and both were impaired relative to Group NC. Group IBO was then given a performance test in which delays and lists were increased stepwise (to delays of 2 min. and to lists lengths of 10). Again, Group IBO's average scores on this performance test did not differ from those of Group ASP, and both were impaired relative to the controls. Although their memory deficits were thus of the same magnitude, Group IBO not only had damage restricted to grey matter but also sustained much smaller perirhinal lesions than Group ASP (about 50 and 90 percent, respectively). It may be concluded that the perirhinal cortex itself is the critical substrate for visual recognition and that there is very little safety factor within the area.